January 28, 2016

Obama, Data & Elections: The Inside Story

Most people who follow politics even in passing are aware that the 2008 and 2012 Obama presidential campaigns were supercharged by dynamic work on paid Facebook campaigns. There’s the lasting mental image of the Democrats’ best and brightest young people sequestered in the basement of the Obama campaign headquarters, slicing and dicing the American population on Facebook -- propelling their man to the presidency, twice.

But that’s pretty much all we know. Even now, people will look back and say something like “Yeah, the Obama campaign team did that really cool stuff on Facebook” without getting into anything specific.

As someone working at a company focused on helping marketers create real-world results (and also as someone who’s spent a lot of time on the backend of Facebook), I perk up my ears whenever the topic is broached. I’m fairly familiar with the advanced tools -- retargeting pixels, lookalike audiences, etc. -- available to marketers with Facebook paid advertising budgets. But the question remained: did the Obama team just outpace the competition with the tools at hand, or did they truly break ground from a data science perspective?

Fortunately, I was able to go directly to the source. Last week I spoke with Andrew Bleeker, who served as Director of Digital Advertising on the 2008 Obama Campaign, and Senior Advisor to the President’s push for reelection in 2012.

He told me that the answer to my question above is — both.

Let’s take a step back for a minute. Voting records generate data that contributes to what is called a “voter file.” Analytics vendors can then use this data to create a predictive model to see what voters are likely to do next Election Day by giving each person a score against his or her file. The people who use this data don't know whom you voted for, just whether you showed up at the polls or not. (Every state has employed a secret ballot since 1891, though the right of voting privacy is not directly protected by the Constitution).

Nevertheless, political campaigns on both sides of the aisle are only concerned with the voter files of only a tiny sliver of the population. That’s because people tend to always vote or always not vote. And the large majority of folks who do go to the polls vote for the same party every time. This leaves only a tiny subset of what Bleeker calls “persuadable people,” preferring that term to the more pervasive yet vague “swing voters.”

“On Facebook and elsewhere we targeted people, not demographics. A lot has been made of ’Soccer Moms’ and ‘Extreme Commuters’ but that’s not how it worked for us,” Bleeker told me. “For example, we targeted the one million people in Ohio our data showed were persuadable. That’s not a demographic.”

Demographics introduce an extra layer of complexity for marketers who want to drive real-world change. Looking at people and what they do in real life -- rather than applying hackneyed, cookie-cutter categories -- made a crucial difference in the Obama campaigns.

But then why didn’t the competition mirror this approach?

In the 2008 general election against John McCain there was simply a talent gap. The Obama campaign generated historic levels of enthusiasm among young people and, since marketing on Facebook was then such a nascent field, those with more personal experience on the platform had a distinct advantage. The McCain campaign just wasn’t as effective targeting persuadable people on Facebook.

By 2012 the talent gap had narrowed significantly, but this time the President and his campaign had a considerable head start. Romney didn’t shake free from Gingrich and Santorum until Super Tuesday on March 6th, only eight months before the general election. Meanwhile the President’s reelection team was continuously building on its data infrastructure from four years earlier, with a digital spending budget of $150 million and a team of 100 full-time employees. Given that knocking on doors is inefficient, it was important to interface with voters on social networks – since they connect online more often than they do in person with neighbors.

For the 2016 general election, Bleeker sees the challenge evolving to approach each persuadable voter more holistically.

“We’re now in a world in which the data available far exceeds the capacity to use it all. Candidates will have to get more dynamic with their creative advertising, they will have to find and nurture people across multiple online channels. The campaigns will have to treat the persuadable voters as people, not just targets.”

The Obama digital team’s process can be instructive for other types of marketing given that it's becoming easier to use historical data to create native models of who’s likely to buy. Attempting to reach a handful of made up demographics might not be in lockstep with the depth of today’s online self-definition capabilities enjoyed by consumers. In its product marketing, Facebook itself admits that custom audiences perform better than demographic targeting does. It helps to cut through the noise created from demographic-centric views -- and get a better look at actual human behavior.

In the end, can we use data to persuade people more effectively online? Yes we can!